Every few years, I teach a post-graduate class which focuses on the structure of argumentation in classic generative syntactic theory. We review important works from the 1960s and 1970s and look at the way in which the foundations of the discipline are built upon one another. Papers from this period are striking in the way that they build tests and probes for specific structural and derivational properties of sentences. For example, arguments from the behavior of idioms, fixed expressions, weather verbs, and embedded rule operations were (and still are) used as probes to distinguish between EQUI (control) constructions and raising constructions (Soames & Perlmutter Reference Soames and Perlmutter1979, see also the review in Davies & Dubinsky Reference Davies and Dubinsky2004). This kind of methodology is still pervasive in generative syntax.

Diagnosing Syntax is a collection of papers that evaluates our modern understanding of the kinds of tests and probes we use to argue for our analyses. Revisiting how we construct these arguments is an important step in understanding how the discipline of syntax has advanced in the past 50 years, and allows us an opportunity to revisit the empirical foundations of the science.

The book opens with an introductory survey by the editors, Lisa Lai-Shen Cheng & Norbert Corver. They pursue a fascinating characterization of the parallels (and lack thereof) between syntactic tests/probes, symptoms and signs, and the related procedures in medical diagnosis. In medicine, diagnosis is ‘the process of identifying the nature (“the hidden structure”) of things on the basis of their properties and characteristics’ (1) and is divided into the examination of signs (objective indications, such as increased heart rate) and symptoms (subjective experiences, such as dizziness or pain). Both of these can be more deeply probed by making use of diagnostic tests, such as an MRI or blood test. The key to successful diagnosis is to be able to evaluate the reliability of the signs and symptoms, particularly in cases where the tests and observations result in evidence that support conflicting hypotheses. For example, how does a doctor distinguish between a fever caused by a bacterial infection and one caused by a virus? What additional steps must be taken to fine-tune the diagnosis?

The charge given to the authors in this volume was to establish what the standard diagnostics for certain syntactic phenomena are (and propose new ones if appropriate) and to evaluate these diagnostics for robustness and reliability. At stake is determining the granularity of the effects that a diagnostic procedure reveals and addressing what happens when tests and symptoms give conflicting diagnoses.

Casting syntactic research in terms of the physician's duties provides a novel way of looking at the everyday business of linguistic analysis, and allows us to step back and consider the bigger picture underlying our basic methodologies. Indeed, the uses of diagnostics in medicine and syntax are often similar. Consider the following quote from David Pesetsky's chapter in this volume: ‘a diagnostic is not a special kind of empirical finding or a component of linguistic theory. Instead the term characterizes a use to which a well-supported older finding can be put when we seek to interpret a new finding’ (153). I think this is true of both clinical medicine and syntactic research. At the same time, the metaphor should not be taken too far. For example, it is not clear to me that medical diagnostics always have the goal of proving a hypothesis, whereas in syntax, that is their explicit purpose. For health care providers, the primary goal of diagnosis is often simply symptom relief, rather than identifying an underlying cause and finding a cure. In the medical world, this is reflected in the distinction between clinical work and bench science. There are no parallels to this distinction in syntactic analysis as far as I can tell.

The editors of the volume identify four classes of diagnostics in generative linguistics: (i) interface diagnostics, which use properties of the PF/LF systems (such as word order, or stress, or scope) to test for syntactic properties; (ii) diagnostics of the syntactic computational system, which give us direct evidence about the form and derivation of sentences – for example, intervention effects in movement or c-command restrictions on anaphor binding; (iii) comparative diagnostics, which give a cross-linguistic or microparametric perspective on language – this kind of research allows us to investigate and reveal the power of diagnostics across and within languages; and, finally, (iv) psycholinguistic and neurolinguistic diagnostics, such as evidence from language acquisition, language processing, and linguistic deficits (such as aphasia). Each section of this book has a chapter devoted to addressing one of these four categories of diagnostics. The sections conclude with a fifth, summarizing chapter. For the most part, I will not discuss these summary chapters here as they recapitulate the content of the other chapters in each part.

The first section, or Part 1 of the book, addresses the diagnostics for head movement. Christer Platzack's contribution, ‘Head movement as a phonological operation’, and Heidi Harley's, ‘Getting morphemes in order: Merger, Affixation and Head Movement’, provide contrasting views on the diagnostics for identifying head movement. Platzack develops a non-syntactic/PF-based account of head movement, relying on a featural stipulation of the place where a head is spelled out in a chain. However, more central to the volume's theme is his identification of four diagnostics for head movement: (a) violations of the extension condition, (b) observation of locality, (c) adherence to relativized minimality, and (d) adherence to the mirror principle. I have to admit that I am not convinced that these are the right diagnostics. Locality and relativized minimality restrictions (diagnostics b and c) are not limited to head movement, but seem to be properties of movement in general. Being a violation of the extension condition is not an empirical test or diagnostic, it is an abstract characterization of a particular theoretical analysis. This leaves adherence to Baker's (Reference Baker1985) mirror principle as the last clear diagnostic in Platzack's palette that singles out head movement. Harley's paper shows that the mirror principle effect, as it is reflected in morpheme order and affixation, need not be diagnostic of head movement either because of the availability of merger-under-adjacency and morpheme-specific linearization principles. As a response to the idea that head movement is really a disguised kind of phrasal movement where all constituents except a head move out of a phrase, and then the remnant phrase dislocates, giving the appearance of head movement, Jochen Zeller, in ‘In defense of head movement: Evidence from Bantu’, shows that there are a variety of intervention effects in Bantu that cannot be explained by a remnant movement approach. In ‘Verb movement to C: From agrammatic aphasia to syntactic analysis’, Naama Friedmann uses evidence from agrammatic aphasia to show that speakers of verb raising languages with different levels of agrammatic impairment show different kinds of errors in the placement of verbs, consistent with the tree-pruning hypothesis. Speakers of Hebrew who have severe impairment affecting their tense systems show significantly less verb movement than those without. This constitutes indirect evidence that verbal heads in Hebrew undergo head movement to the T node.

The next section of the book concentrates on phrasal movement of all kinds. Personally, I would have preferred to see separate sections on different kinds of phrasal movement (A-movement, A-bar movement, scrambling, etc.) because the diagnostics for each type of movement are, almost by definition, going to be distinct. This said, David Pesetsky's, ‘Phrasal movement and its discontents: Diseases and diagnoses’, provides a nice ‘diagnostic manual’ for phrasal movement, by pulling out critical features of all kinds of movement. A moved item has properties of both its underlying position and its surface position and movement is a means to resolve any conflicting requirements of those two positions. He then shows how this abstraction can be applied to show the reliability of a wide variety of diagnostic tests of different kinds of (overt) movement. The articles by Winfried Lechner and Maria Polinsky & Eric Potsdam (‘Diagnosing covert movement: The Duke of York and reconstruction’ and ‘Diagnosing covert A-movement’, respectively) address the challenging issue of demonstrating the existence of movement in cases where there is no surface evidence. Lechner develops two new tests for covert wh-movement. Polinsky & Potsdam compare DP in situ in Adyghe and Russian and give evidence from intervention effects, control and binding to show that Adyghe has covert DP movement but Russian does not. Hamida Demirdache's chapter, ‘Arguments for long distance movement in long distance questions in child language’, surveys the literature on child language and long distance questions. She looks in particular at how established diagnostics for long distance movement prove that children have successive cyclic movement.

The first two sections of the book focus on how to diagnose a particular kind of construction. The third section in the book flips the question on its head. The contributions here look at one particular diagnostic – morphological agreement – and investigate the extent to which tracks the abstract feature checking operation of Agree (Chomsky Reference Chomsky and Kenstowicz2001). The consensus of the first two papers in this section is that morphological agreement is not a reliable diagnostic for Agree. ‘The syntactic relations behind agreement’, by Sandra Chung, demonstrates convincingly that while morphological agreement overlaps with the relations typically associated with Agree, it does not coincide with it. The author does this by showing that morphological agreement can overlap with other syntactic relations, such as specifier–head relations and adjacency. She also shows that morphological agreement does not respect the locality restrictions imposed by Agree. Focusing on gender agreement, Ora Matushansky, in ‘Gender confusion’, comes to a similar conclusion for cases involving ‘mixed’ agreement patterns. The well known phenomenon in Arabic, where subject–verb orders result in a fuller agreement pattern than verb–subject orders is addressed in Jamal Ouhalla's ‘Agreement unified: Arabic’. The author argues against the usual account of such constructions (such as Mohammad Reference Mohammad1989), in which the richer pattern of agreement is due to a specifier–head relationship in subject–verb orders. He uses evidence from agreement inside DPs to show that the partial agreement pattern is due to an independently motivated rule of agreement impoverishment and again is not indicative of the Agree operation. The paper by Maria Teresa Guasti, ‘Agreement in the production of subject and object wh-questions’, has a different take. The author suggests that processing and production problems involved in non-canonical, long distance Agree configurations in Italian wh-constructions are reflected in the unusual agreement patterns exhibited by these phenomena.

The section on anaphora has a mix of perspectives. Elena Anagnostopoulou & Martin Everaert, in ‘Identifying anaphoric dependencies’, treat anaphoric relations as the target of diagnosis. They provide a rich new set of data on Greek o idhios, which shows properties of both pronouns and anaphors using a number of diagnostic properties such as ability to be free, restrictions on what can c-command the anaphor, split antecedence, and strict/sloppy interpretations. The authors show that of the range of possible diagnostics for anaphora, deictic reference and c-command restrictions are the most reliable of tests. Norvin Richards' paper, ‘Tagalog anaphora’, uses anaphora as a probe to investigate the difference between scrambling and topicalization in Tagalog. Richards claims that topicalization is a covert operation in the LF component, but scrambling is an operation that happens mid-phase between the LF-spellout and PF-spellout operations. Christopher Tancredi's ‘Condition B’ makes the provocative proposal that by using phase theory, one can dispense with condition B. The idea is interesting but its tie-in to the volume's theme is less clear. Sergey Avrutin & Sergio Baauw's ‘A processing view on agrammatism’ is a manifesto for the view that agrammatism is largely a processing deficit rather than a grammatical one. This would suggest that data from patients with aphasia should not be treated as tests for syntactic theory. Having this paper in the section on anaphora seemed out of place, however, because while it did touch on pronominal interpretation, the paper had a much wider empirical scope and a goal that seemed to be more about the correct explanation for agrammatism than about diagnostics of anaphora.

The first chapter in the final section (on ellipsis) is Jason Merchant's ‘Polarity items under ellipsis’, which focuses on the nature of the identity condition on ellipsis sites, particularly when there is an apparent mismatch between the licensor and a polarity item in an omitted structure (as in, for example, John didn't see anyone but Mary did). Merchant proposes a late insertion approach, where the content of polarity items is a function of surface licensing by a negative head. The content of Susanne Winkler's chapter, ‘Syntactic diagnostics for extraction of focus from ellipsis site’, is self-evident from the title (which, as an aside, is ungrammatical for this Canadian native speaker of English). She identifies a set of diagnostics based on focus theory that allow one to characterize the behavior of two different types of ellipsis (contrastive vs. non-contrastive) with respect to extraction. Explaining the acceptability of mismatch ellipsis (such as This tuna could have been eaten by the cat, but he decided not to) is the goal of Lyn Frazier's ‘A recycling approach to processing ellipsis’. She gives a processing account of the variation in acceptability of these sentences, claiming that the more repairs a processor has to do, the more unacceptable the sentence. This, she claims, interferes with some of the tests that have been proposed to identify ellipsis constructions. Jeroen van Craenenbroeck & Anikó Lipták, in ‘What sluicing can do, what it can't and in which language’, present evidence from Hungarian that sluicing is not limited to wh-constructions. Instead, under the right circumstances in the right languages, sluicing may allow for a licensing remnant to be a focused element. They then use this refined notion of sluicing as a probe to investigate the different left-edge positions for wh- and focused constituents in languages that allow multiple wh-movement.

Despite the fact that the book is quite long, I finished it wishing there had been some more chapters on diagnostics in other areas of syntax, such as control theory or constituent structure tests. Nevertheless, the articles in Diagnosing Syntax provide a substantive and important contribution to our understanding of what we can (and cannot) use as reliable tests when we're trying to reveal the abstract properties of the sentence structure and its derivation.